adds model preference support for mcp server sampling requests

[bkrabach]

Example:

from mcp.types import ModelPreferences
from mcp_extensions import send_sampling_request

# Set model preferences
model_preferences = ModelPreferences(
    # Can use hints to prefer models with specific names
    # hints=[
    #     ModelHint(
    #         # Prefer models where name starts with `name` value (so `o3` would _include_ `o3-mini`)')
    #         name="gpt-4o",
    #     )
    # ],
    # Setting speed priority to 1 to choose a faster model, like gpt-4o
    speedPriority=1,
    # If needing a reasoning model, set intelligence priority to 1 instead
    # intelligencePriority=1,
)

sampling_result = await send_sampling_request(
        fastmcp_server_context=ctx,
        system_prompt="<your prompt>",
        messages=messages,
        model_preferences=model_preferences,
        max_tokens=1024,
    )